OpenAI defends firing of 3 AI safety researchers, cites ‘significant breach of trust’

OpenAI has defended the firing of three AI safety researchers, Mikita Balesni, Tomek Korbak, and Jasmine Wang, citing a "significant breach of trust". The researchers, who raised concerns about their firings potentially discouraging other employees from speaking up about AI safety, argue that advanced AI cannot be developed safely without collaborative work and independent experts.

Korbak revealed that OpenAI told him he was fired for how he communicated with the external safety organization METR, which they partnered with to investigate the Hugging Face hack. Balesni and Wang also alleged they were dismissed for communicating with METR, denying any wrongdoing. OpenAI maintains that internal investigations found a breach of policies regarding sensitive information handling, but denies the firings were related to safety concerns.

The researchers argue that their firings set a dangerous precedent, discouraging employees from raising safety concerns or collaborating with outside safety groups.